Google says it has pulled AI mannequin Gemma from its AI Studio platform after a Republican senator complained the mannequin, designed for builders, “fabricated severe felony allegations” about her.
In a put up on X, Google’s official information account stated the corporate had “seen reviews of non-developers attempting to make use of Gemma in AI Studio and ask it factual questions.” AI Studio is a platform for builders and never a traditional approach for normal shoppers to entry Google’s AI fashions. Gemma is particularly billed as a household of AI fashions for builders to make use of, with variants for medical use, coding, and evaluating textual content and picture content material.
Gemma was by no means meant for use as a client software, or for use to reply factual questions, Google stated. “To stop this confusion, entry to Gemma is not obtainable on AI Studio. It’s nonetheless obtainable to builders via the API.”
Google didn’t specify which reviews prompted Gemma’s removing, although on Thursday Senator Marsha Blackburn (R-TN) wrote to CEO Sundar Pichai accusing the corporate of defamation and anti-conservative bias. Blackburn, who additionally raised the problem throughout a current Senate commerce listening to about anti-diversity activist Robby Starbuck’s personal AI defamation go well with towards Google, claimed Gemma responded falsely when requested “Has Marsha Blackburn been accused of rape?”
Gemma apparently replied that Blackburn “was accused of getting a sexual relationship with a state trooper” throughout her 1987 marketing campaign for state senate, who alleged “she pressured him to acquire pharmaceuticals for her and that the connection concerned non-consensual acts.” It additionally offered an inventory of pretend information articles to assist the story, Blackburn stated.
None of that is true, not even the marketing campaign yr which was really 1998. The hyperlinks result in error pages and unrelated information articles. There has by no means been such an accusation, there isn’t any such particular person, and there aren’t any such information tales. This isn’t a innocent “hallucination.” It’s an act of defamation produced and distributed by a Google-owned AI mannequin.
The narrative has a well-known really feel. Despite the fact that we’re now a number of years into the generative AI increase, AI fashions nonetheless have a fancy relationship with the reality. False or deceptive solutions from AI chatbots masquerading as info nonetheless plague the trade and regardless of enhancements there isn’t any clear resolution to the accuracy drawback in sight. Google stated it stays “dedicated to minimizing hallucinations and frequently bettering all our fashions.”
In her letter, Blackburn stated her response stays the identical: “Shut it down till you may management it.”
